Fear of a black president: Mississippi edition

i-3eb03977d9930d4f23c17c682c281121-mississippiobama.jpgThe map to the left shows the counties which voted for Obama (blue) and McCain (red) in the 2008 election. The blue counties are part of the Black Belt, the area where blacks are a majority of the population because of the economic concentration of cotton culture during the 19th and 20th centuries. The McCain Belt, those counties where John McCain beat George W. Bush, is getting some press, but obviously it is interesting to wonder about areas where large black populations which increased turnout are likely masking the shift of the white vote for John McCain. I have already shown on a state-by-state basis where the white vote shifted toward the Democrats in 2008, and where it shifted toward the Republicans. Though the average white vote budged only a bit, there is important regional structure which is being masked by aggregating all this information.

The political scientist Larry Bartels reaffirms my basic point:

However, there is a good deal of circumstantial evidence suggesting that racial resentment eroded Obama's support among white voters. His gains relative to Kerry were significantly smaller in states with large numbers of African-Americans--a pattern disguised in the overall vote totals by his strong support among African-Americans themselves. In the former Confederacy he gained only slightly over Kerry among white voters, despite making big gains in two key swing states, North Carolina and Virginia. The only states in the country in which he lost more than a point or two of white support were Louisiana, Alabama, Arkansas, and Mississippi.

You can see this in the exit polls. They rather straightforwardly illustrate that Obama won a smaller percentage of the white vote than Kerry in many states in the South. But I decided to look at it a different way: I plotted the percentage of whites in each county and their vote percentages for Barack Obama and John Kerry in Mississippi. This is more precise than an exit poll because votes are votes, and the Census counts everyone. So here is that chart:


As you can see, as the percentage of whites increases the proportion voting for the Democrat decreases. This is no surprise. In this part of the South the Republicans are the white party, and the Democrats are the black party. But look at the slopes: it is higher in 2008 than 2004. That means that the white vote for Obama was lower and the black vote for him higher. The "ends" of the line swiveled so that the slope tilted up. If I assume that this is modeled by a linear regression, the r-squared for 2008 is 0.93 and for 2004 it is 0.76. If I assume that this is a logistic curve the r-squared for 2008 is 0.90 and for 2004 is 0.81. The r-squared can be interpreted so that one can say that ~90% of the variation on a state wide level in voting for Democrats could be explained by the proportion who were white in the counties in 2008. In 2004 the equivalent value would be ~75-80%. Both of these values are high, but, it suggests that the exit polls are right, and racial polarization has increased over the 4 years. Whiteness is a nearly perfect proxy for the Republican presidential vote now, and blackness the Democratic vote. And these associations have increased between 2004 and 2008.

But it isn't as if Mississippi is one homogeneous state with one culture. Among whites there are those who are basically in Appalachia, while others who live in the lowland South. So I decided to constrain the chart to counties with various intervals of the white %. Here is a chart of only the majority black counties:


Here are majority white counties:

And here are counties where 62% or more are white, which is the proportion in Mississippi which is white (so these are the whiter counties in the state):

It looks to me that what you're seeing are two things:

1) Increased black turnout in counties where many blacks weren't voting. I think the black turnout went up nationally by about 10% (that is, from 11% to 12% of the electorate), but there might be regional disparities in terms of where that 10% is coming from.

2) Some white people in mostly white counties switched from Democrats to Republicans.

I assume in majority black counties you're pretty much as polarized as you can get; very few white Democrats and black Republicans. In the overwhelmingly white counties there's less polarization because the fact that there are few black people means that white Democrats aren't a member of the black party, at least locally. But it looks like the emergence of Obama changed things even in very white counties, where there is a discernible drop off in white support, with obviously far less compensation from the relatively small black minority turning out a higher rate. If you're wondering about what that one outlier on the chart which is far away from the trendlines, it's Alcorn county in Mississippi's far north. It's not a data entry error, that was my first thought....

Here's the raw data:

White Obama Kerry
Adams 43 59 55
Alcorn 87.8 27 61
Amite 56.6 44 42
Attala 52.5 42 38
Benton 65 48 53
Bolivar 33.4 67 64
Calhoun 70.1 35 35
Carroll 65.2 34 34
Chickasaw 57.5 50 49
Choctaw 68.5 35 33
Claiborne 14.6 85 82
Clarke 65.5 37 32
Clay 42.5 59 52
Coahoma 25.3 73 64
Copiah 48.4 53 43
Covington 63.2 41 38
Desoto 78.8 30 27
Forrest 62.5 43 38
Franklin 63.2 37 35
George 90.1 16 22
Greene 72.5 24 27
Grenada 57.4 24 41
Hancock 90.4 21.2 29
Harrison 71.8 36.6 36
Hinds 32.4 70 60
Holmes 18.2 82 76
Humphreys 25.6 71 64
Issaquena 39.3 61 53
Itawamba 92.5 21 29
Jackson 74.7 32 30
Jasper 46.9 55 51
Jeff Davis 41.7 61 51
Jefferson 13.5 87 81
Jones 71.6 30 28
Kemper 37.9 63 53
Lafayette 72.3 43 40
Lamar 83.5 21 19
Lauderdale 57.7 40 34
Lawrence 67.7 37 36
Leake County 55.6 44 40
Lee 72.3 34 33
Leflore 28.2 68 62
Lincoln 68.9 34 31
Lowndes 54.6 58 38
Madison 60.6 42 35
Marion 66.2 34 33
Marshall 50.3 59 59
Monroe 68.4 41 40
Montgomery 53.8 46 45
Neshoba 64.2 26 25
Newton 65.2 32 27
Noxubee 29.1 76 71
Okitbbeha 58.2 50 43
Panola 51.4 52 49
Pearl River 86.3 19 23
Perry 76.3 27 25
Pike 49.3 51 47
Pontotoc 84.5 23 23
Prentiss 84.9 28 32
Quitman 30 67 60
Rankin 79.4 23 20
Scott 59.6 43 37
Sharkey 30.6 68 50
Simpson 64.6 39 32
Smith 75.8 24 21
Stone 79.5 27 27
Sunflower 27.1 70 63
Tallahatchie 39 59 55
Tate 68.4 39 39
Tippah 81.3 27 33
Tishomingo 95.4 23 34
Tunica 26.6 76 69
Union 84 24 27
Walthall 54.7 44 39
Warren 51.9 49 41
Washington 31.3 67 59
Wayne 61.5 39 36
Webster 78.1 25 26
Wilkinson 30.1 69 64
Winston 54.6 45 53
Yalobusha 60.2 46 44
Yazoo 44.1 57 45

More like this

I post some data analysis over at my other weblog. For example, today I looked at the relationship between food stamp usage and unemployment. The Census makes a lot of county-level data available, though it's often slapdash and disorganized. But using R I've constructed many data sets including…
In 1996 Bill Clinton won with 49% of the vote vs. 41% for Bob Dole. The New York Times now allows you to compare county-by-county outcomes across two elections between all presidential years between 1992 and 2008. I think 1996 is the most analogous to Barack Obama's victory yesterday, so I want…
John Emerson points me to some interesting data crunching over at Open Left. The diarist, "dreaminonempty," is analyzing the past few years' election results against demographic variables. What's there not to like? Though I do think the perspective is a bit too The Emerging Democratic Majority. Yes…
I've made my first stab at a prediction for the electoral college outcome for the US Presidential race, 2016. I use a roughly similar methodology as I did to accurately predict most of the Democratic primaries. However, since primaries are different from a general, the methodology had to be…

Razib, I don't know if you follow the Horserace blog (http://www.realclearpolitics.com/horseraceblog), but Jay Cost examined very similar questions vis-a-vis Obama vs. Hillary this spring. The question was a bit different -- what is Obama's voting coalition, and how does Hillary break into it, but there were regional differences that appeared at that point and that still appear in the Obama/McCain match-up. The series is worth reading if you haven't.